Performance & Scalability Characterization. By Richard Tibbetts Co-Founder and Chief Architect, StreamBase Systems, Inc.



Similar documents
StreamBase White Paper StreamBase versus Native Threading. By Richard Tibbetts Co-Founder and Chief Technology Officer, StreamBase Systems, Inc.

Case Study: BlueCrest Capital Management Ltd.

TIBCO StreamBase High Availability Deploy Mission-Critical TIBCO StreamBase Applications in a Fault Tolerant Configuration

StreamBase White Paper Real-Time Profit & Loss

Removing Performance Bottlenecks in Databases with Red Hat Enterprise Linux and Violin Memory Flash Storage Arrays. Red Hat Performance Engineering

Low Latency Market Data and Ticker Plant Technology. SpryWare.

What can DDS do for You? Learn how dynamic publish-subscribe messaging can improve the flexibility and scalability of your applications.

Millennium SOR. Fast. Flexible. Robust.

White Paper. Recording Server Virtualization

VirtualCenter Database Performance for Microsoft SQL Server 2005 VirtualCenter 2.5

Virtuoso and Database Scalability

TIBCO Live Datamart: Push-Based Real-Time Analytics

FX Trading: A Changing Landscape

FX Trading and Technology Trends 2012

Data Center Storage Solutions

Real Time Performance Dashboard for SOA Web Services ORION SOA

How To Store Data On An Ocora Nosql Database On A Flash Memory Device On A Microsoft Flash Memory 2 (Iomemory)

Informatica Ultra Messaging SMX Shared-Memory Transport

Tableau Server 7.0 scalability

The IBM Cognos Platform for Enterprise Business Intelligence

Switching Architectures for Cloud Network Designs

PRODUCTIVITY IN FOCUS PERFORMANCE MANAGEMENT SOFTWARE FOR MAILROOM AND SCANNING OPERATIONS

Four Keys to Successful Multicore Optimization for Machine Vision. White Paper

Oracle Database Scalability in VMware ESX VMware ESX 3.5

TIBCO StreamBase High Availability Deploy Mission-Critical TIBCO StreamBase Applications in a Fault Tolerant Configuration

I N T E R S Y S T E M S W H I T E P A P E R INTERSYSTEMS CACHÉ AS AN ALTERNATIVE TO IN-MEMORY DATABASES. David Kaaret InterSystems Corporation

HOW INTERSYSTEMS TECHNOLOGY ENABLES BUSINESS INTELLIGENCE SOLUTIONS

Achieving Nanosecond Latency Between Applications with IPC Shared Memory Messaging

Improve Business Productivity and User Experience with a SanDisk Powered SQL Server 2014 In-Memory OLTP Database

Benefits of multi-core, time-critical, high volume, real-time data analysis for trading and risk management

WITH A FUSION POWERED SQL SERVER 2014 IN-MEMORY OLTP DATABASE

Improving Grid Processing Efficiency through Compute-Data Confluence

Dell Microsoft Business Intelligence and Data Warehousing Reference Configuration Performance Results Phase III

Green HPC - Dynamic Power Management in HPC

QLogic 16Gb Gen 5 Fibre Channel for Database and Business Analytics

How To Improve Your Communication With An Informatica Ultra Messaging Streaming Edition

StreamStorage: High-throughput and Scalable Storage Technology for Streaming Data

Testing Big data is one of the biggest

SafePeak Case Study: Large Microsoft SharePoint with SafePeak

Lambda Architecture. Near Real-Time Big Data Analytics Using Hadoop. January Website:

Mission-Critical Java. An Oracle White Paper Updated October 2008

A Comparison of Oracle Performance on Physical and VMware Servers

Low-latency market data delivery to seize competitive advantage. WebSphere Front Office for Financial Markets: Fast, scalable access to market data.

SQL Server Consolidation Using Cisco Unified Computing System and Microsoft Hyper-V

Case Study - I. Industry: Social Networking Website Technology : J2EE AJAX, Spring, MySQL, Weblogic, Windows Server 2008.

VMware vcenter 4.0 Database Performance for Microsoft SQL Server 2008

I N T E R S Y S T E M S W H I T E P A P E R F O R F I N A N C I A L SERVICES EXECUTIVES. Deploying an elastic Data Fabric with caché

BigMemory & Hybris : Working together to improve the e-commerce customer experience

Software-defined Storage at the Speed of Flash

White Paper. How Streaming Data Analytics Enables Real-Time Decisions

Improving Scalability for Citrix Presentation Server

Enhance visibility into and control over software projects IBM Rational change and release management software

Unprecedented Performance and Scalability Demonstrated For Meter Data Management:

TekSouth Fights US Air Force Data Center Sprawl with iomemory

IBM Cognos 10: Enhancing query processing performance for IBM Netezza appliances

Recommendations for Performance Benchmarking

Real-time Big Data Analytics with Storm

CASE STUDY: Oracle TimesTen In-Memory Database and Shared Disk HA Implementation at Instance level. -ORACLE TIMESTEN 11gR1

Reference Architecture, Requirements, Gaps, Roles

Cloud Building Blocks: Paving the Road

Monitoring can be as simple as waiting

Unified network traffic monitoring for physical and VMware environments

A Comparison of Oracle Performance on Physical and VMware Servers

Condusiv s V-locity Server Boosts Performance of SQL Server 2012 by 55%

WHITE PAPER. GemFireTM on Intel for Financial Services (GIFS) Reference Architecture for Trading Systems. Overview

SOLUTION BRIEF. TIBCO StreamBase for Algorithmic Trading

Upgrading to Microsoft SQL Server 2008 R2 from Microsoft SQL Server 2008, SQL Server 2005, and SQL Server 2000

SMB Direct for SQL Server and Private Cloud

Scalability and Performance Report - Analyzer 2007

Tableau Server Scalability Explained

Getting Real Real Time Data Integration Patterns and Architectures

Leveraging EMC Fully Automated Storage Tiering (FAST) and FAST Cache for SQL Server Enterprise Deployments

SUN ORACLE EXADATA STORAGE SERVER

Trading. Theory and Practice

be architected pool of servers reliability and

Rackspace Cloud Databases and Container-based Virtualization

FTEN Trade Control Gateways

Dell One Identity Manager Scalability and Performance

Enhance Service Delivery and Accelerate Financial Applications with Consolidated Market Data

Milestone Solution Partner IT Infrastructure MTP Certification Report Scality RING Software-Defined Storage

Boost Database Performance with the Cisco UCS Storage Accelerator

SAP HANA From Relational OLAP Database to Big Data Infrastructure

Accelerating Server Storage Performance on Lenovo ThinkServer

KPI, OEE AND DOWNTIME ANALYTICS. An ICONICS Whitepaper

Company & Solution Profile

Aras Innovator 10 Scalability Benchmark Methodology and Performance Results

The Evolution of Microsoft SQL Server: The right time for Violin flash Memory Arrays

Virtualizing SQL Server 2008 Using EMC VNX Series and Microsoft Windows Server 2008 R2 Hyper-V. Reference Architecture

Inform IT Enterprise Historian. The Industrial IT Solution for Information Management

Solving the Hypervisor Network I/O Bottleneck Solarflare Virtualization Acceleration

Transcription:

Performance & Scalability Characterization By Richard Tibbetts Co-Founder and Chief Architect, StreamBase Systems, Inc.

Motivation for Performance and Scalability in a CEP Engine CEP engines can be applied to a wide range of problems facing capital markets firms. In each of these problems, it is the CEP engine that performs the application s critical processing and therefore sits at the center of the flow of real-time and historical data streams. A low perfomring CEP engine can be the bottleneck of any application, unacceptably reducing the performance. Therefore, performance and scalability are important criteria when evaluating a CEP engine. This report measures the performance and scalability of an Equities Alternative Trading System (ATS) application using the StreamBase CEP engine and illustrates why leading global investment banks, hedge funds and other capital markets firms find StreamBase is the best performing CEP engine for the job. We measure throughput, latency, and scalability of the application. We conclude that on our baseline quad-core machine, the StreamBase ATS handles 140,000 input messages per second with latency of 86 microseconds. Scaling to an eight-core AMD machine, the StreamBase ATS processes 245,000 input messages per second, with 71 microseconds of latency, demonstrating a scaling factor of 0.875 across multi-core architectures. Choosing an Application In order to develop a representative measurement of the performance and scalability of a CEP engine, we must first identify the key attributes of the measurement: Full Application: CEP engines are used to build whole applications, and it is better to measure the performance and scalability of an entire application rather than focusing on the performance of individual pieces of that application. Heterogeneous Messages: The application being measured must process different types of messages to be fully representative of a common trading application. State Management: The application being measured should require the storage and maintenance of state, not just simple data transformation, in order to exercise the capabilities of the CEP engine in this area. Metrics for Performance Performance of CEP engines is multi-dimensional. In order to fully characterize performance, three things should be measured: Throughput: How many messages per second can the application process? This defines the raw speed of the application, how much data it can process in a steady state, as well as how it will deal with load spikes. 2

Latency: What is the time between the engine receiving a message, and the results of that message being sent out? Latency is critical to trading operations, because a late answer is often a wrong answer, as market conditions change. Scalability: How do throughput and latency change when the application is moved to different hardware, such as a machine with more processing cores. Data volumes are always increasing, and it is important that a CEP application be scalable so that it does not have to be rewritten in the future. Equities ATS Benchmark The application measured in this report is based on an Equities Alternative Trading System (ATS). The ATS is responsible for holding and crossing orders inside a broker/dealer. This application might run inside any major broker, handling both internal and client order flow. The ATS handles heterogeneous messages with both market data and multiple kinds of orders. It also maintains state, in the form of a book of pending orders to be matched against. The message inputs to an Equities ATS include: Market Data: In order to perform legal crosses, market data must be monitored and NBBO maintained Orders: Client order flow must pass through the ATS, in order to create liquidity. Orders are either matched against existing orders in ATS, or passed to downstream markets for execution. Indications of Interest (IOI): When an internal desk is looking for liquidity, they can place an indication of interest into the ATS, which will rest there waiting for a cross, rather than being sent to the downstream market The message outputs from an Equities ATS include: Order and IOI Echo: Orders and IOIs are also sent to downstream systems for record keeping Order Ack to Client: Upstream clients must receive acknowledgements when their order is resting in the ATS Fill Messages: When an order fills or partially fills, the ATS must also send notifications to the client The StreamBase Application Diagram A below shows the StreamBase Equities ATS application. Input streams on the left receive market data, orders, and indications of interest. These data are normalized before being delivered to the core of the application, where matching occurs. An order book is maintained in memory of pending orders. New orders are matched against the pending orders. If there is a match, a fill message is created, which notifies the client and triggers the booking of a trade. Other parts of the application are responsible for calculating monitoring statistics, such as throughput, order resting time, and fill rates, and for logging actions to create an audit trail. 3

Diagram A: StreamBase Equities ATS Application Hardware Specifications We use a 4-core Dell PowerEdge 1360 server and an 8-core AMD whitebox server to measure the performance and scalability of the StreamBase Equities ATS application. Chart A shows the hardware specifications of these multi-core machines. Chart A: Hardware Specifications 4

Performance and Scalability Results Overall Results We measured throughput, latency, and scalability across the two hardware configurations. On our baseline quad-core machine, the StreamBase ATS handles 140,000 input messages per second with latency of 86 microseconds. Scaling to an eight-core AMD machine, the StreamBase ATS processes 245,000 input messages per second, with 71 microseconds of latency, demonstrating a scaling factor of 0.875 across multi-core architectures. Throughput Results Executing on the standard 4-core machine, the StreamBase CEP engine was able to maintain in excess of 140,000 messages per second. This input was composed of 118,600 market data updates per second and 21,900 orders per second. On the 8-core machine, throughput was 245,400 input messages per second. The breakdown was 207,800 market data updates per second and 37,600 orders per second. Graph A shows the input and output results of the Equities ATS. Graph A: Equities ATS Input/Output Results Latency Results On the 4-core machine, average latency was 86 microseconds. On the 8-core machine, average latency was 71 microseconds. 5

Scalability Results In cooperation with AMD, we demonstrate the scaling of the StreamBase Equities ATS application when moving from a 4-core server to an 8-core server. We take advantage of the parallelism in the application to run portions of the load in parallel. Graph B below shows that the AMD 8-core machine supports higher throughput and greater parallelism with lower latency. Graph B: Scaling of the Equities ATS Benchmark 6

About Richard Tibbetts Richard Tibbetts is co-founder and Chief Architect at StreamBase Systems. Throughout his career, Tibbetts has concentrated on the design and development of database technologies, while also offering specialization in program analysis and distributed computing. Tibbetts created the Linear Road Benchmark, the database management industry s guide for comparing the performance of real-time streaming databases with alternative systems, like relational databases. In addition, he has been heavily involved with a number of academic database projects, including Aurora at Brown University, and Medusa at the Massachusetts Institute of Technology (MIT). In 2003, the research from these two projects later led to the founding and commercialization of StreamBase s Stream Processing Engine. As a co-founder and Chief Architect at StreamBase, Tibbetts is directly involved in architecting the foundation for the next-generation textual language based on Structured Query Language (SQL). The objective of this work is to apply the benefits of SQL for stored data to real-time, transitory data. Tibbetts is also charged with global responsibilities in furthering new StreamBase capabilities through platform and technical innovation. Tibbetts earned both his bachelor of science in computer science and engineering and his master s of engineering degree at MIT. He completed his graduate research under the direction of Dr. Mike Stonebraker. About StreamBase StreamBase Systems, Inc, a leader in high-performance Complex Event Processing (CEP), provides software for rapidly building systems that analyze and act on real-time streaming data for instantaneous decision-making. StreamBase s Event Processing Platform(tm) combines a rapid application development environment, an ultra low-latency highthroughput event server, and the broadest connectivity to real-time and historical data and leading EMS/OMS software platforms. Six of the top ten Wall Street investment banks and three of the top five hedge funds use StreamBase to power mission-critical applications to increase revenue, lower costs, and reduce risk. It is also used by government agencies for highly specialized intelligence work. The company is headquartered in Lexington, Massachusetts with European offices in London. For more information, visit www.streambase.com. Corporate Headquarters 181 Spring Street Lexington, MA 02421 +1 (866) 787-6227 New York Office 845 Third Avenue, 6th Floor New York, NY 10022 +1 (866) 787-6227 Virginia Office 11921 Freedom Drive, Suite 550 Reston, VA 20190 +1 (443) 472-0767 European Headquarters 60 Cannon Street London, EC4N 6NP +44 (0) 20 7002 1095 www.streambase.com 2009 StreamBase Systems, Inc. All rights reserved. StreamBase, StreamBase Studio, and When Now Means Right Now are trademarks of StreamBase Systems, Inc. All other trademarks are the property of their respective owners. 7